[Script Info]
Title: 
[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
Dialogue: 0,0:00:00.11,0:00:14.52,Default,,0000,0000,0000,,{\i1}34C3 preroll music{\i0}\N
Dialogue: 0,0:00:14.52,0:00:15.97,Default,,0000,0000,0000,,Herald: The next speaker is born and
Dialogue: 0,0:00:15.97,0:00:21.65,Default,,0000,0000,0000,,raised in Germany. He lives and works as a\NPhD student in Canada as a member of a
Dialogue: 0,0:00:21.65,0:00:27.32,Default,,0000,0000,0000,,research group on extremist politics in\Ndemocratic systems and he'll give us an
Dialogue: 0,0:00:27.32,0:00:32.64,Default,,0000,0000,0000,,insight into the public discourse in\NGermany focused on the so called
Dialogue: 0,0:00:32.64,0:00:37.75,Default,,0000,0000,0000,,"Alternative für Deutschland". Please\Nwelcome Alexander Beyer.
Dialogue: 0,0:00:37.75,0:00:45.25,Default,,0000,0000,0000,,{\i1}applause{\i0}
Dialogue: 0,0:00:45.25,0:00:53.12,Default,,0000,0000,0000,,Beyer: Thank you very much. Thank you\Npeople, for showing up in the Saal Borg,
Dialogue: 0,0:00:53.12,0:00:58.98,Default,,0000,0000,0000,,thank you, the internet, for watching, a\Nvery big thank you for the organizers, for
Dialogue: 0,0:00:58.98,0:01:03.100,Default,,0000,0000,0000,,giving me the opportunity to, to give this\Nlittle talk. Yeah, my name is Alexander
Dialogue: 0,0:01:03.100,0:01:11.02,Default,,0000,0000,0000,,Beyer and everywhere I went this winter, I\Ndidn't have to wear a winter jacket
Dialogue: 0,0:01:11.02,0:01:15.51,Default,,0000,0000,0000,,because the temperatures were very mild\Nand I will tell you in a minute why that
Dialogue: 0,0:01:15.51,0:01:19.60,Default,,0000,0000,0000,,matters.\NAs I already said, I'm a member of a
Dialogue: 0,0:01:19.60,0:01:27.79,Default,,0000,0000,0000,,research group in Vancouver, where we look\Nat what happens, how extremist parties and
Dialogue: 0,0:01:27.79,0:01:33.94,Default,,0000,0000,0000,,politics fare in democratic systems and\Nwe decided to focus this research project
Dialogue: 0,0:01:33.94,0:01:40.23,Default,,0000,0000,0000,,on the fascinating - for researchers\Nfascinating - case of Germany and asking
Dialogue: 0,0:01:40.23,0:01:46.64,Default,,0000,0000,0000,,the questions, if we can point fingers and\Nis it a viable, a valid
Dialogue: 0,0:01:46.64,0:01:53.01,Default,,0000,0000,0000,,judgment to say, the media, the\Nmedia is to blame for the rise of the AfD?
Dialogue: 0,0:01:53.01,0:01:58.79,Default,,0000,0000,0000,,For anyone who, who decided at the end of\N2017 that they would spend most of 2016 in
Dialogue: 0,0:01:58.79,0:02:02.59,Default,,0000,0000,0000,,hibernation - which seemed like a pretty\Ngood idea at the time - I will give a
Dialogue: 0,0:02:02.59,0:02:09.77,Default,,0000,0000,0000,,quick rundown what happened: So, we had\Nelection in September and the domino
Dialogue: 0,0:02:09.77,0:02:16.52,Default,,0000,0000,0000,,piece that was Germany fell. Domino piece\Nin a sense, that all around in Europe far-
Dialogue: 0,0:02:16.52,0:02:22.66,Default,,0000,0000,0000,,right parties had considerable success in\Nthe past, in the recent past, and Germany
Dialogue: 0,0:02:22.66,0:02:26.90,Default,,0000,0000,0000,,was the sort of last stalwart in\NCentral Europe, where a far-right party did
Dialogue: 0,0:02:26.90,0:02:35.76,Default,,0000,0000,0000,,not get at the government, this\Nhappened in September and it did not only
Dialogue: 0,0:02:35.76,0:02:40.75,Default,,0000,0000,0000,,get into government, er, into Parliament, it\Nalso, the way that it looks like now, it
Dialogue: 0,0:02:40.75,0:02:43.76,Default,,0000,0000,0000,,might become the official leader of the\Nopposition.
Dialogue: 0,0:02:43.76,0:02:49.92,Default,,0000,0000,0000,,So, when these results came in, pundits\Nwere really, really quick to call the
Dialogue: 0,0:02:49.92,0:02:55.48,Default,,0000,0000,0000,,shots.\NThe, the dominating sentiment was, that it
Dialogue: 0,0:02:55.48,0:03:00.93,Default,,0000,0000,0000,,was the media's fault: They took the\Npositions of the AfD and gave
Dialogue: 0,0:03:00.93,0:03:08.47,Default,,0000,0000,0000,,disproportionate amounts of coverage to\Nthis far-right extremist party. And this
Dialogue: 0,0:03:08.47,0:03:16.35,Default,,0000,0000,0000,,sentiment had a lot of truthiness to it.\NSo, it had a lot of: "Yeah, sure, I can see
Dialogue: 0,0:03:16.35,0:03:21.38,Default,,0000,0000,0000,,why", right, everyone that opened his\Nnewspaper or opened a news website,
Dialogue: 0,0:03:21.38,0:03:24.76,Default,,0000,0000,0000,,stories about the AfD seemed, or about\Nanything.. about something that's related
Dialogue: 0,0:03:24.76,0:03:31.84,Default,,0000,0000,0000,,to the AfD, seemed to dominate coverage.\NThis went along with a little bit of a
Dialogue: 0,0:03:31.84,0:03:39.76,Default,,0000,0000,0000,,felt truth, a truth that was perceived by\Npeople, about how the campaigning season
Dialogue: 0,0:03:39.76,0:03:43.89,Default,,0000,0000,0000,,was a lot of season and not a lot of\Ncampaigning, despite Martin Schultz's best
Dialogue: 0,0:03:43.89,0:03:48.16,Default,,0000,0000,0000,,efforts.\NA whole lot of sunshine, but not a lot of
Dialogue: 0,0:03:48.16,0:03:55.14,Default,,0000,0000,0000,,conflict and this was something that then\Nwas perceived to be very, uh, well, I don't
Dialogue: 0,0:03:55.14,0:04:04.15,Default,,0000,0000,0000,,want to say very skillfully, but somehow\Nfilled by the AfD and and the topics that
Dialogue: 0,0:04:04.15,0:04:10.52,Default,,0000,0000,0000,,are of concern to this party.\NSo what are we doing today here? First
Dialogue: 0,0:04:10.52,0:04:17.64,Default,,0000,0000,0000,,off, I'm a political scientist by trade\Nand political scientists like theory. I
Dialogue: 0,0:04:17.64,0:04:22.43,Default,,0000,0000,0000,,know that this is an event, where [I] figure\Nyou might not at the forefront of
Dialogue: 0,0:04:22.43,0:04:27.50,Default,,0000,0000,0000,,everyone's minds, but it is for me,\Nbecause talking and arguing to political
Dialogue: 0,0:04:27.50,0:04:32.55,Default,,0000,0000,0000,,scientists about theory is kind of like\Nmud-wrestling with a pig: you do that for
Dialogue: 0,0:04:32.55,0:04:39.27,Default,,0000,0000,0000,,two or three hours and then you realize,\Noh, this pig actually enjoys this. So I'll
Dialogue: 0,0:04:39.27,0:04:45.96,Default,,0000,0000,0000,,be sort of, I have one slide on what we,\Nwhat previous theories would suggest
Dialogue: 0,0:04:45.96,0:04:50.69,Default,,0000,0000,0000,,has, have happened and how it could have\Nhappened and then I'll show you what kind
Dialogue: 0,0:04:50.69,0:04:56.52,Default,,0000,0000,0000,,of data we have collected, to\Nsystematically answer this question and
Dialogue: 0,0:04:56.52,0:05:02.04,Default,,0000,0000,0000,,talk about public discourse in Germany and\Nthen to the meat and potatoes of the talk,
Dialogue: 0,0:05:02.04,0:05:09.09,Default,,0000,0000,0000,,about how the campaign unfolded in the\Nmedia and I will than, to end I will show
Dialogue: 0,0:05:09.09,0:05:15.17,Default,,0000,0000,0000,,some more data, that is a bit different,\Nthat paints a picture on why this election
Dialogue: 0,0:05:15.17,0:05:21.80,Default,,0000,0000,0000,,was a special election and why it was sort\Nof a perfect storm of an election for a
Dialogue: 0,0:05:21.80,0:05:29.92,Default,,0000,0000,0000,,far-right party and why this actually\Nmakes us claim that the media could be
Dialogue: 0,0:05:29.92,0:05:35.53,Default,,0000,0000,0000,,said to have behaved pretty reasonable. As\Na little teaser.
Dialogue: 0,0:05:35.53,0:05:44.26,Default,,0000,0000,0000,,OK. Theory. One slide. Two possible\Nmechanisms of media effects. There's this
Dialogue: 0,0:05:44.26,0:05:52.19,Default,,0000,0000,0000,,normative, very endearing and wonderful\Nidea, that if you read something, that
Dialogue: 0,0:05:52.19,0:05:59.85,Default,,0000,0000,0000,,someone carefully crafts and he or she\Nconstructs an argument, that is well
Dialogue: 0,0:05:59.85,0:06:05.100,Default,,0000,0000,0000,,written, well made, you read this, you take\Nit in, you're persuaded by that,
Dialogue: 0,0:06:05.100,0:06:13.41,Default,,0000,0000,0000,,regardless of what this argument is. 60\Nyears of media research suggests, that this
Dialogue: 0,0:06:13.41,0:06:17.33,Default,,0000,0000,0000,,doesn't happen.\NPre-existing opinions are extremely
Dialogue: 0,0:06:17.33,0:06:21.93,Default,,0000,0000,0000,,difficult to change in each and every\Nsingle one of us, even though we're likely
Dialogue: 0,0:06:21.93,0:06:28.09,Default,,0000,0000,0000,,to admit that "No, no sure, I'm a rational\Nthinker, I take standpoints if they're
Dialogue: 0,0:06:28.09,0:06:35.77,Default,,0000,0000,0000,,convincing to me and I internalize them.",\Nbut it's not how this works.
Dialogue: 0,0:06:35.77,0:06:42.49,Default,,0000,0000,0000,,The second possible effect, and the one\Nthat will be of of concern to us today at
Dialogue: 0,0:06:42.49,0:06:50.56,Default,,0000,0000,0000,,the core of the presentation, is something\Nthat's called Priming. So, the media can't
Dialogue: 0,0:06:50.56,0:06:55.01,Default,,0000,0000,0000,,tell people what to think, it can't\Npersuade people independently of the
Dialogue: 0,0:06:55.01,0:07:00.64,Default,,0000,0000,0000,,previous opinions that people have, but\Nit's really, really successful in telling
Dialogue: 0,0:07:00.64,0:07:05.86,Default,,0000,0000,0000,,people, what to think about. It's super\Ngood, the media is super good, reading
Dialogue: 0,0:07:05.86,0:07:11.50,Default,,0000,0000,0000,,something is very effective in bringing\Nsomething to the front of your mind.
Dialogue: 0,0:07:11.50,0:07:18.27,Default,,0000,0000,0000,,And here I.. here I can tell you, why I\Ntold you about my my choice of attire in
Dialogue: 0,0:07:18.27,0:07:22.06,Default,,0000,0000,0000,,winter. A vast majority of you probably\Nthought when I said this "'Oh, I didn't
Dialogue: 0,0:07:22.06,0:07:27.91,Default,,0000,0000,0000,,have to wear a winter jacket', wow, what's..\Nwho is this guy?" But maybe a few of you
Dialogue: 0,0:07:27.91,0:07:31.51,Default,,0000,0000,0000,,thought "Yeah, sure it was pretty mild,\Nthat's climate change."
Dialogue: 0,0:07:31.51,0:07:37.97,Default,,0000,0000,0000,,So, without naming the issue, I..\Nthere's a chance, that I primed a few of
Dialogue: 0,0:07:37.97,0:07:43.17,Default,,0000,0000,0000,,you, to consider climate change and pull\Nthat in your frontal lobe, at the front of
Dialogue: 0,0:07:43.17,0:07:49.49,Default,,0000,0000,0000,,your mind.\NThis is, this is important, this is a.. the
Dialogue: 0,0:07:49.49,0:07:57.05,Default,,0000,0000,0000,,central thing, that we have to consider if\Nwe ask, if the media wrote up a party like
Dialogue: 0,0:07:57.05,0:08:03.26,Default,,0000,0000,0000,,the AfD.\NAlso important to consider here is, that
Dialogue: 0,0:08:03.26,0:08:10.78,Default,,0000,0000,0000,,priming is easier - or there's an indirect\Neffect of priming as well - where a topic
Dialogue: 0,0:08:10.78,0:08:17.03,Default,,0000,0000,0000,,that is owned by a specific party, that's\Nthe thing that then favors the party
Dialogue: 0,0:08:17.03,0:08:20.75,Default,,0000,0000,0000,,subsequently.\NSo, if the media writes a lot about
Dialogue: 0,0:08:20.75,0:08:28.59,Default,,0000,0000,0000,,refugees, a xenophobic far-right party,\Nthat has this problem of refugees at the
Dialogue: 0,0:08:28.59,0:08:38.36,Default,,0000,0000,0000,,core of their agenda, will reap in benefits\Nin our minds, in that it's agenda will be,
Dialogue: 0,0:08:38.36,0:08:44.91,Default,,0000,0000,0000,,will fall on fertile ground.\NSo far the theory, that's all. So, what did
Dialogue: 0,0:08:44.91,0:08:51.04,Default,,0000,0000,0000,,we do based on this theory? We collected\Ndata, lots of data, we have.. we understand this
Dialogue: 0,0:08:51.04,0:08:56.74,Default,,0000,0000,0000,,data.. we understand this text, that we\Ncollected to be data and we use natural
Dialogue: 0,0:08:56.74,0:09:00.67,Default,,0000,0000,0000,,language processing to analyze that.\NNatural Language Processing basically
Dialogue: 0,0:09:00.67,0:09:04.90,Default,,0000,0000,0000,,means, that we're giving language to a\Ncomputer that wasn't written specifically
Dialogue: 0,0:09:04.90,0:09:12.10,Default,,0000,0000,0000,,to be understood by a computer and try to\Nextract meaningful analysis based on what
Dialogue: 0,0:09:12.10,0:09:17.83,Default,,0000,0000,0000,,the computer is doing with this.\NSo, we used some sifting methods to
Dialogue: 0,0:09:17.83,0:09:25.77,Default,,0000,0000,0000,,collect about 8.500 articles from four\Ncentral German news websites: Focus, Bild,
Dialogue: 0,0:09:25.77,0:09:31.86,Default,,0000,0000,0000,,Welt and Spiegel. And we have.. that\Nresults in a unique data set, that, to our
Dialogue: 0,0:09:31.86,0:09:39.25,Default,,0000,0000,0000,,knowledge, no one else has. If so, please\Nreach out to us.
Dialogue: 0,0:09:39.25,0:09:45.94,Default,,0000,0000,0000,,And this was so unique, that it deserves at\Nleast six fire emojis. It was also pretty
Dialogue: 0,0:09:45.94,0:09:51.35,Default,,0000,0000,0000,,exciting because, that was pretty cheap. We\Nwere two people that were mainly concerned
Dialogue: 0,0:09:51.35,0:09:56.64,Default,,0000,0000,0000,,with collecting this data and I don't want\Nto, I don't want to calculate my hourly
Dialogue: 0,0:09:56.64,0:10:03.75,Default,,0000,0000,0000,,wage, but it was almost done with no\Nfinancial expense. And this is cool,
Dialogue: 0,0:10:03.75,0:10:08.74,Default,,0000,0000,0000,,because we're social scientists, we're\Nfaced with this problem, with this very
Dialogue: 0,0:10:08.74,0:10:15.90,Default,,0000,0000,0000,,interesting case of Germany sort of\Nfalling in line, very delayed, with lots of
Dialogue: 0,0:10:15.90,0:10:20.43,Default,,0000,0000,0000,,other countries around it - in terms of\Nthe far-right party getting their seats in
Dialogue: 0,0:10:20.43,0:10:25.23,Default,,0000,0000,0000,,Parliament and we can use methods that are\Navailable to us, if we're sort of like
Dialogue: 0,0:10:25.23,0:10:27.62,Default,,0000,0000,0000,,sitting down and reading our stack\Noverflow and sort of teaching those
Dialogue: 0,0:10:27.62,0:10:32.04,Default,,0000,0000,0000,,methods to us, to systematically try to\Nanswer this question.
Dialogue: 0,0:10:32.04,0:10:38.68,Default,,0000,0000,0000,,Let's dive right in. The share of party\Nmentions in online news. So, what we did
Dialogue: 0,0:10:38.68,0:10:45.98,Default,,0000,0000,0000,,for each day, we calculated what the total\Nnumber of mentioned political actors is.
Dialogue: 0,0:10:45.98,0:10:51.99,Default,,0000,0000,0000,,We did that based on word lists, that we\Ncarefully crafted, that included candidate's
Dialogue: 0,0:10:51.99,0:10:56.77,Default,,0000,0000,0000,,names and party abbreviations and party\Nnames and things like "Kanzlerin" and
Dialogue: 0,0:10:56.77,0:11:05.67,Default,,0000,0000,0000,,"Kanzlerkandiat" for the CDU/CSU and the\NSPD respectively and we let that thing rip
Dialogue: 0,0:11:05.67,0:11:12.70,Default,,0000,0000,0000,,through our little rscript that we have.\NSo, the average of mentions of each party
Dialogue: 0,0:11:12.70,0:11:19.27,Default,,0000,0000,0000,,over the course of the campaign looks\Nsomething like this: Between July 1st and
Dialogue: 0,0:11:19.27,0:11:25.93,Default,,0000,0000,0000,,September 24th - that's the time frame\Nthat we concentrated on - we see a clear
Dialogue: 0,0:11:25.93,0:11:31.79,Default,,0000,0000,0000,,incumbency bonus, the "Kanzlerbonus" the\N"Kanzlerinbonus" for the CDU/CSU, social
Dialogue: 0,0:11:31.79,0:11:40.10,Default,,0000,0000,0000,,democrats high twenties, and the AfD at\N10.7 %. Here we might say at smaller
Dialogue: 0,0:11:40.10,0:11:44.45,Default,,0000,0000,0000,,parties.. a little note to the green and\Nthe left, so with this dictionary method is
Dialogue: 0,0:11:44.45,0:11:48.99,Default,,0000,0000,0000,,kind of tricky, because we can't say: "Oh yeah,\Nwell we just gonna count every occurrence
Dialogue: 0,0:11:48.99,0:11:55.38,Default,,0000,0000,0000,,of Grüne and Linke" for the Green Party and\Nthe Left Party, because then we get stuff
Dialogue: 0,0:11:55.38,0:12:00.53,Default,,0000,0000,0000,,like 'the green banana' and 'the left hand'\Nthat is counted for them. So that's why
Dialogue: 0,0:12:00.53,0:12:03.96,Default,,0000,0000,0000,,here we're only using candidates' names.\NThat's why they probably.. they sort of
Dialogue: 0,0:12:03.96,0:12:09.33,Default,,0000,0000,0000,,underperform. But for our purpose of\Ntalking about why the AFD got favored by
Dialogue: 0,0:12:09.33,0:12:14.36,Default,,0000,0000,0000,,the media, we're sort of letting that drop\Nunder the table. So, the story here is
Dialogue: 0,0:12:14.36,0:12:21.74,Default,,0000,0000,0000,,over the course of the campaign, 10.7 % of\Nmentions were happening that mentioned the
Dialogue: 0,0:12:21.74,0:12:28.75,Default,,0000,0000,0000,,AFD, basically. Case closed. Right, AFD got\N12.7 % in the election. That doesn't really
Dialogue: 0,0:12:28.75,0:12:36.04,Default,,0000,0000,0000,,sound like it was favored by the media.\NAnd a few of you might know this analysis
Dialogue: 0,0:12:36.04,0:12:42.51,Default,,0000,0000,0000,,from a blog post, that me and Constanze\NKurz wrote for Netzpolitik, sort of like 45
Dialogue: 0,0:12:42.51,0:12:47.08,Default,,0000,0000,0000,,seconds after the election, when we worked\Non truncated data. And we also focused on
Dialogue: 0,0:12:47.08,0:12:54.72,Default,,0000,0000,0000,,print media and this is sort of what this\Ngraph looked like, that we based our
Dialogue: 0,0:12:54.72,0:12:59.48,Default,,0000,0000,0000,,conclusion on. AFD didn't really get any\Ndisproportionate amount of coverage. It
Dialogue: 0,0:12:59.48,0:13:02.48,Default,,0000,0000,0000,,actually is in the.. in the last week of the\Ncampaign.. last weeks of the campaign
Dialogue: 0,0:13:02.48,0:13:10.41,Default,,0000,0000,0000,,actually is outperformed by the FDP.\NScience is the current state of airing, or
Dialogue: 0,0:13:10.41,0:13:18.37,Default,,0000,0000,0000,,the.. so, now that we have better data in\Nterms of online news data this whole story
Dialogue: 0,0:13:18.37,0:13:23.09,Default,,0000,0000,0000,,looks a bit different. If we take the\Naverage over the whole course of the
Dialogue: 0,0:13:23.09,0:13:29.14,Default,,0000,0000,0000,,campaign and actually have it shown to\Nus.. Stay by the day - This is what I want
Dialogue: 0,0:13:29.14,0:13:36.03,Default,,0000,0000,0000,,to focus on now.\NSo just looking at the sort of tail end of
Dialogue: 0,0:13:36.03,0:13:41.09,Default,,0000,0000,0000,,this all the way to the right, when we get\Nclose to the election date, the order of
Dialogue: 0,0:13:41.09,0:13:46.75,Default,,0000,0000,0000,,this is surprisingly close to the actual\Nelection results. The parties actually do
Dialogue: 0,0:13:46.75,0:13:58.38,Default,,0000,0000,0000,,get in the order, that they came out of\Nthe election. But we do see a little curve
Dialogue: 0,0:13:58.38,0:14:04.90,Default,,0000,0000,0000,,that gets closer to a curve that should be\Nbigger. And this is where the.. well, I
Dialogue: 0,0:14:04.90,0:14:09.83,Default,,0000,0000,0000,,don't want to say magic, but this is\Nwhere the interesting stuff lies. So let's
Dialogue: 0,0:14:09.83,0:14:16.64,Default,,0000,0000,0000,,look at the curves one after the other.\NThe CDU/CSU, as you would expect, as the
Dialogue: 0,0:14:16.64,0:14:21.59,Default,,0000,0000,0000,,incumbent anything that is remotely\Npolitical in domestic and international
Dialogue: 0,0:14:21.59,0:14:26.57,Default,,0000,0000,0000,,politics, will score mentions for the\NChancellor and the CDU/CSU. That's why
Dialogue: 0,0:14:26.57,0:14:30.48,Default,,0000,0000,0000,,this curve is considerably higher than the\Nothers, but we do see a downward tendency
Dialogue: 0,0:14:30.48,0:14:34.94,Default,,0000,0000,0000,,the closer we get to the campaign, when\Ncampaign coverage shifted from the
Dialogue: 0,0:14:34.94,0:14:41.22,Default,,0000,0000,0000,,incumbent to the competitors. Especially\Nthe underdog competitors. Which is kind..
Dialogue: 0,0:14:41.22,0:14:47.67,Default,,0000,0000,0000,,that's a bad transfer to the SPD\Nnow, but if we look at the curve of the
Dialogue: 0,0:14:47.67,0:14:52.83,Default,,0000,0000,0000,,Social Democratic Party, there's a slight\Nbump around August and Martin Schulz
Dialogue: 0,0:14:52.83,0:14:59.75,Default,,0000,0000,0000,,really tried to drive home this issue of\Njustice as the central campaign promise
Dialogue: 0,0:14:59.75,0:15:02.87,Default,,0000,0000,0000,,and there's another little slight\Nhumper on September 1st, begin of
Dialogue: 0,0:15:02.87,0:15:08.43,Default,,0000,0000,0000,,September, when the televised debate\Nhappened, but the overall trend is pretty
Dialogue: 0,0:15:08.43,0:15:13.46,Default,,0000,0000,0000,,linear, doesn't seem to be, if we would\Njust smooth this plot out to be a straight
Dialogue: 0,0:15:13.46,0:15:20.43,Default,,0000,0000,0000,,line, it probably would be pretty much\Nhorizontal. Not so for the AFD. So
Dialogue: 0,0:15:20.43,0:15:25.11,Default,,0000,0000,0000,,remember, over the course of the campaign\Nthey got 10.7 % on average of mentions
Dialogue: 0,0:15:25.11,0:15:28.93,Default,,0000,0000,0000,,{\i1}drinks from his bottle{\i0}\NAnd that's true. If we calculate an
Dialogue: 0,0:15:28.93,0:15:35.53,Default,,0000,0000,0000,,average of that, of course, this looks like\Nit scores considerably lower than the two
Dialogue: 0,0:15:35.53,0:15:42.10,Default,,0000,0000,0000,,major parties. But something happens in\Nlate August and all of a sudden this party
Dialogue: 0,0:15:42.10,0:15:46.70,Default,,0000,0000,0000,,gets actually close to the Social\NDemocrats. It like.. starting in late
Dialogue: 0,0:15:46.70,0:15:52.69,Default,,0000,0000,0000,,August, the tendency becomes one, that is\Npretty considerably upwards. And if we
Dialogue: 0,0:15:52.69,0:15:57.67,Default,,0000,0000,0000,,take the average of only the two last\Nweeks before the election, we get to a
Dialogue: 0,0:15:57.67,0:16:05.03,Default,,0000,0000,0000,,number of 19.6 of all mentions are talking\Nabout the AFD there. Which is something, if
Dialogue: 0,0:16:05.03,0:16:11.02,Default,,0000,0000,0000,,you think about the mechanisms of priming,\Nthose are short term effects. We're
Dialogue: 0,0:16:11.02,0:16:14.39,Default,,0000,0000,0000,,looking for things that happen over a\Nshort term or have an effect in a pretty
Dialogue: 0,0:16:14.39,0:16:20.57,Default,,0000,0000,0000,,short term. So this is something that is\Nextremely important. At the beginning of
Dialogue: 0,0:16:20.57,0:16:24.65,Default,,0000,0000,0000,,this time frame, where the plot becomes\Nsomething that has a trend that shows
Dialogue: 0,0:16:24.65,0:16:30.67,Default,,0000,0000,0000,,upwards, around, like, August 28th - where\Nthat first little mountain.. first little
Dialogue: 0,0:16:30.67,0:16:35.93,Default,,0000,0000,0000,,summit occurs - two things happened: one, a\Nrefugee boat capsized in the
Dialogue: 0,0:16:35.93,0:16:41.17,Default,,0000,0000,0000,,Mediterranean. An event that we've sadly\Nand.. have to see terrifyingly often and
Dialogue: 0,0:16:41.17,0:16:45.20,Default,,0000,0000,0000,,one of the people died. And the second\Nthing that happened was, that Alexander
Dialogue: 0,0:16:45.20,0:16:50.99,Default,,0000,0000,0000,,Gauland in an interview claimed, that a\NGerman politician should be dumped in
Dialogue: 0,0:16:50.99,0:17:02.55,Default,,0000,0000,0000,,Anatolia. And it's interesting, if you talk\Nabout.. if you extract the topics, that
Dialogue: 0,0:17:02.55,0:17:07.51,Default,,0000,0000,0000,,are covered in relation to the AFD before\Nand after this moment.
Dialogue: 0,0:17:07.51,0:17:15.33,Default,,0000,0000,0000,,Before this August 28th, it's a lot about\NAlice Weidel writing emails where it turns
Dialogue: 0,0:17:15.33,0:17:20.38,Default,,0000,0000,0000,,out, she's not the public persona that she\Nclaims she is and it's a lot about
Dialogue: 0,0:17:20.38,0:17:26.41,Default,,0000,0000,0000,,internal rifts of this far-right party.\NThe internal tensions between the super
Dialogue: 0,0:17:26.41,0:17:35.03,Default,,0000,0000,0000,,far right wing and the far right or right\Nwing-wing and afterwards, there's a
Dialogue: 0,0:17:35.03,0:17:41.61,Default,,0000,0000,0000,,surprising amount of citations of this "Oh\Nwe're gonna... we should dump this person
Dialogue: 0,0:17:41.61,0:17:47.18,Default,,0000,0000,0000,,in another country." So that's something\Nthat indicates, that this strategy of sort
Dialogue: 0,0:17:47.18,0:17:56.36,Default,,0000,0000,0000,,of provoking a scandal paid off. But\Nlet's.. before we get into that, let's
Dialogue: 0,0:17:56.36,0:17:59.74,Default,,0000,0000,0000,,look into the topics that were covered\Nover the course of the campaign. We did
Dialogue: 0,0:17:59.74,0:18:05.27,Default,,0000,0000,0000,,the same thing, we developed topic\Ndictionaries with keywords for each
Dialogue: 0,0:18:05.27,0:18:09.53,Default,,0000,0000,0000,,category and we let our script read\Nthrough all the data and count
Dialogue: 0,0:18:09.53,0:18:18.52,Default,,0000,0000,0000,,occurrences. So looking at this, we see a\Nsort of band there in the 10% range,
Dialogue: 0,0:18:18.52,0:18:24.64,Default,,0000,0000,0000,,where it's all a colourful rainbow, where\Nthe topics don't really diverge from each
Dialogue: 0,0:18:24.64,0:18:29.24,Default,,0000,0000,0000,,other, except for that\Ntopic of domestic security, which
Dialogue: 0,0:18:29.24,0:18:34.15,Default,,0000,0000,0000,,is there at the low end of the range. But\Nwe do have one topic, that stands out quite
Dialogue: 0,0:18:34.15,0:18:39.70,Default,,0000,0000,0000,,considerably in the early months of the\NWahlkampfsommer, which is European Union,
Dialogue: 0,0:18:39.70,0:18:47.01,Default,,0000,0000,0000,,generally European Union topic. This is\Nbecause on July 1st Helmut Kohl, the
Dialogue: 0,0:18:47.01,0:18:51.95,Default,,0000,0000,0000,,eternal Chancellor, get the first European\Nact of state and a lot of things were
Dialogue: 0,0:18:51.95,0:18:55.91,Default,,0000,0000,0000,,written about his legacy in terms of the\NEuropean Union and lots of people showed
Dialogue: 0,0:18:55.91,0:19:00.25,Default,,0000,0000,0000,,up from Strasbourg and Brussels and paid\Ntheir respects. This is why this topic
Dialogue: 0,0:19:00.25,0:19:05.99,Default,,0000,0000,0000,,seems like.. or this is, why this topic comes\Nin as strong as it does here. Now the
Dialogue: 0,0:19:05.99,0:19:10.59,Default,,0000,0000,0000,,topic that has a sort of unusual curve\Nhere on our graph, is the topic of the
Dialogue: 0,0:19:10.59,0:19:17.68,Default,,0000,0000,0000,,environment. Our dictionaries that we\Ndeveloped were topical and so what causes
Dialogue: 0,0:19:17.68,0:19:26.09,Default,,0000,0000,0000,,this steep, steep summit there in early\NAugust, is the Dieselgipfel.. there, the
Dialogue: 0,0:19:26.09,0:19:31.70,Default,,0000,0000,0000,,Diesel Summit, where German car\Nmanufacturers try to sort of get out of
Dialogue: 0,0:19:31.70,0:19:38.67,Default,,0000,0000,0000,,the fact, that they basically ripped off\Ncustomers with selling cars that emitted
Dialogue: 0,0:19:38.67,0:19:44.25,Default,,0000,0000,0000,,toxic amounts of poisonous gas and dust.\NThis is why this is extremely important in
Dialogue: 0,0:19:44.25,0:19:49.00,Default,,0000,0000,0000,,the high.. in the low 40 % range in\Nearly August. But afterwards the trend
Dialogue: 0,0:19:49.00,0:19:55.37,Default,,0000,0000,0000,,line points steeply down.\NA topic that was pretty consistent over
Dialogue: 0,0:19:55.37,0:20:01.58,Default,,0000,0000,0000,,the course of the campaign in its overall\Ndynamic or at the sort of.. not the
Dialogue: 0,0:20:01.58,0:20:06.46,Default,,0000,0000,0000,,overall dynamic, but the role that it\Nplayed, is the topic of immigration. And
Dialogue: 0,0:20:06.46,0:20:13.91,Default,,0000,0000,0000,,immigration means migration and refugees\Nin our case here. And now, thinking about
Dialogue: 0,0:20:13.91,0:20:19.98,Default,,0000,0000,0000,,what that means in relation to our theory\Non priming, we would think that sure,
Dialogue: 0,0:20:19.98,0:20:25.58,Default,,0000,0000,0000,,that's a topic that is owned by the AFD.\NIt's, like, it's super tightly connected to
Dialogue: 0,0:20:25.58,0:20:34.36,Default,,0000,0000,0000,,that party's rise. So, this is something\Nthat does favour a far-right party like
Dialogue: 0,0:20:34.36,0:20:40.89,Default,,0000,0000,0000,,those are, like it is. But we can do\Na sort of more systematic investigation
Dialogue: 0,0:20:40.89,0:20:50.71,Default,,0000,0000,0000,,into this. So, this graph shows you the\Npoles: each dot represents polling results
Dialogue: 0,0:20:50.71,0:20:55.90,Default,,0000,0000,0000,,for the AFD and the line is the average\Nout of those polls, again over the course
Dialogue: 0,0:20:55.90,0:20:59.99,Default,,0000,0000,0000,,of the of the time frame that we surveyed.\NPretty much constant
Dialogue: 0,0:20:59.99,0:21:03.84,Default,,0000,0000,0000,,until mid-August, and all of a sudden we have\Nincreasing variance and we have a tendency,
Dialogue: 0,0:21:03.84,0:21:11.10,Default,,0000,0000,0000,,a trend line, that points upwards. And now,\Nthis is where the heart of the story lies:
Dialogue: 0,0:21:11.10,0:21:19.28,Default,,0000,0000,0000,,is this, is this dependent on the\Nmentions that the AFD got in the media?
Dialogue: 0,0:21:19.28,0:21:23.01,Default,,0000,0000,0000,,There's the orange line, now we have a\Nsort of we have a different, a different
Dialogue: 0,0:21:23.01,0:21:26.47,Default,,0000,0000,0000,,scale of our graph, that's why it looks\Nway more nervous than in the bigger one,
Dialogue: 0,0:21:26.47,0:21:34.67,Default,,0000,0000,0000,,that we had. Difficult to say. If you have\Ndata like this, time serious data, you
Dialogue: 0,0:21:34.67,0:21:38.85,Default,,0000,0000,0000,,actually want to get rid of trends, in\Nterms of what the analysis should be like.
Dialogue: 0,0:21:38.85,0:21:44.49,Default,,0000,0000,0000,,So one way to do this in a graphic\Nrepresentation, is by not showing the
Dialogue: 0,0:21:44.49,0:21:49.59,Default,,0000,0000,0000,,absolute values and how they develop, but\Nonly showing the change from day to day
Dialogue: 0,0:21:49.59,0:21:57.49,Default,,0000,0000,0000,,and plotting that. This is what this graph\Ndoes. So here, these two lines dance around
Dialogue: 0,0:21:57.49,0:22:02.39,Default,,0000,0000,0000,,the zero mark, because - especially the blue\None, where it's the polling results - there
Dialogue: 0,0:22:02.39,0:22:06.14,Default,,0000,0000,0000,,wasn't a lot of variation from day to day.\NIt's in incremental steps that the curve
Dialogue: 0,0:22:06.14,0:22:13.37,Default,,0000,0000,0000,,points up and down. It gets a bit more, a\Nbit higher in variance around the.. after
Dialogue: 0,0:22:13.37,0:22:20.29,Default,,0000,0000,0000,,the mid of August. And whereas the AFD\Nmentions in the media, they stay, well, they
Dialogue: 0,0:22:20.29,0:22:26.87,Default,,0000,0000,0000,,stay rich in variance. Hard to tell, if\Nanything systematic is there. You would
Dialogue: 0,0:22:26.87,0:22:33.39,Default,,0000,0000,0000,,think that after the sort of first, 3rd\Nof August, those, those lines are
Dialogue: 0,0:22:33.39,0:22:38.89,Default,,0000,0000,0000,,connected. We ran an analysis - a vector\Nautoregression model, time series
Dialogue: 0,0:22:38.89,0:22:45.16,Default,,0000,0000,0000,,statistics - we couldn't find any\Nsystematic relation in a timeframe, that
Dialogue: 0,0:22:45.16,0:22:48.79,Default,,0000,0000,0000,,made sense for our theory on priming.\NWhich is a few days that we're looking
Dialogue: 0,0:22:48.79,0:22:53.77,Default,,0000,0000,0000,,for. So if you talk about time series, we\Ntalk about lag and lead, and so you try to
Dialogue: 0,0:22:53.77,0:22:58.64,Default,,0000,0000,0000,,connect a data point that is further down\Nthe line with a data point that is, that is
Dialogue: 0,0:22:58.64,0:23:07.55,Default,,0000,0000,0000,,not as far down the line, and nothing of\Nstatistical significance showed up here.
Dialogue: 0,0:23:07.55,0:23:10.96,Default,,0000,0000,0000,,And this kind of stumped us - we thought,\Nright, when we looked at this there was
Dialogue: 0,0:23:10.96,0:23:18.51,Default,,0000,0000,0000,,something. That we.. we sort of took a step\Nback and we considered another possibility
Dialogue: 0,0:23:18.51,0:23:29.40,Default,,0000,0000,0000,,to.. as why to.. as to why the media reported\Nas they did. Did the media just give
Dialogue: 0,0:23:29.40,0:23:37.66,Default,,0000,0000,0000,,the people what the people wanted?\NAnd here is why I want to talk
Dialogue: 0,0:23:37.66,0:23:43.47,Default,,0000,0000,0000,,to you about why this was a\Nspecial election. This graph - I adapted
Dialogue: 0,0:23:43.47,0:23:48.19,Default,,0000,0000,0000,,this graph from the Berliner Morgenpost\Nand they based it on on surveys conducted
Dialogue: 0,0:23:48.19,0:23:54.38,Default,,0000,0000,0000,,by Infratest dimap on to the.. the data\NI didn't have any access to them. But this
Dialogue: 0,0:23:54.38,0:24:00.26,Default,,0000,0000,0000,,impressively shows, why there was a special\Nelection. In five out of the six preceding
Dialogue: 0,0:24:00.26,0:24:05.35,Default,,0000,0000,0000,,elections, employment was the topic that\Nwas on top of people's minds, when they
Dialogue: 0,0:24:05.35,0:24:10.39,Default,,0000,0000,0000,,made the decision in terms of which party\Nto vote for. And employment means
Dialogue: 0,0:24:10.39,0:24:17.75,Default,,0000,0000,0000,,unemployment. In 2017, with unemployment\Nbeing at record lows, and after 2015
Dialogue: 0,0:24:17.75,0:24:23.72,Default,,0000,0000,0000,,having.. or having a Syrian civil war still\Ngoing on, we're having ...
Dialogue: 0,0:24:23.72,0:24:32.30,Default,,0000,0000,0000,,refugees come into.. into Western and into\NEurope, immigration jumps on out as the.. as
Dialogue: 0,0:24:32.30,0:24:38.68,Default,,0000,0000,0000,,the topic that was the most important for\Npeople. And here, if we if we look at also
Dialogue: 0,0:24:38.68,0:24:44.76,Default,,0000,0000,0000,,the topics that are further down the\Nimportant scale for voters, those are all
Dialogue: 0,0:24:44.76,0:24:50.83,Default,,0000,0000,0000,,topics, where one could conceivably think\Nthat those can be spun in a way that they
Dialogue: 0,0:24:50.83,0:24:58.04,Default,,0000,0000,0000,,are connected to this refugee situation.\NSocial injustice, economic injustice, that's
Dialogue: 0,0:24:58.04,0:25:02.34,Default,,0000,0000,0000,,something that a party like the AFD can\Nvery effectively turn into an idea on
Dialogue: 0,0:25:02.34,0:25:10.56,Default,,0000,0000,0000,,group based conflict: "It's us versus them".\NSame with pensions: "Oh, those people come
Dialogue: 0,0:25:10.56,0:25:15.32,Default,,0000,0000,0000,,here to take our jobs and our money, and\Nespecially from the old people. From our
Dialogue: 0,0:25:15.32,0:25:26.46,Default,,0000,0000,0000,,elderly. So 2017, die Bundestagswahl 2017,\Nis a special case, if we consider it
Dialogue: 0,0:25:26.46,0:25:33.29,Default,,0000,0000,0000,,compared to other parties. So now having\Nthis situation where we find that it's,
Dialogue: 0,0:25:33.29,0:25:36.58,Default,,0000,0000,0000,,it's something that basically never\Nhappened in recent history and in Germany
Dialogue: 0,0:25:36.58,0:25:42.73,Default,,0000,0000,0000,,before in terms of what, what made people\Ndecide at the polls. We wondered, OK,
Dialogue: 0,0:25:42.73,0:25:48.39,Default,,0000,0000,0000,,well, is there a way to more accurately\Nmeasure this demand side of things, this
Dialogue: 0,0:25:48.39,0:25:55.29,Default,,0000,0000,0000,,this need for information for.. of voters.\NAnd what better way there is to measure
Dialogue: 0,0:25:55.29,0:26:03.51,Default,,0000,0000,0000,,some.. to measure the salience in the\Npopulation than to look at Google queries?
Dialogue: 0,0:26:03.51,0:26:10.26,Default,,0000,0000,0000,,So we collected Google Trends data - more\Nspecifically, the Google searches on
Dialogue: 0,0:26:10.26,0:26:13.98,Default,,0000,0000,0000,,refugees, 'Flüchtlinge', \Ngeneral term,
Dialogue: 0,0:26:13.98,0:26:17.78,Default,,0000,0000,0000,,and again, here's this..\Nthis way to even out a trend
Dialogue: 0,0:26:17.78,0:26:25.19,Default,,0000,0000,0000,,line - this is the daily change in\Nhow this topic developed. And if we put
Dialogue: 0,0:26:25.19,0:26:31.77,Default,,0000,0000,0000,,our daily change of AFD mentions over\Nthat, we do see that there's something
Dialogue: 0,0:26:31.77,0:26:40.37,Default,,0000,0000,0000,,there. There's some sort of systematic\Nrelationship. And then, crunching these
Dialogue: 0,0:26:40.37,0:26:44.70,Default,,0000,0000,0000,,numbers and putting them again through a\Nvector autoregression model, we come to
Dialogue: 0,0:26:44.70,0:26:52.97,Default,,0000,0000,0000,,the conclusion, that with a lag of only one\Nday, Google searches for refugees actually
Dialogue: 0,0:26:52.97,0:26:58.56,Default,,0000,0000,0000,,lead AFD mentions in the media. So, if on\NTuesday a higher number of people in
Dialogue: 0,0:26:58.56,0:27:06.76,Default,,0000,0000,0000,,Germany googled "Refugees", on Wednesday, the\NAFD was mentioned more often than the day
Dialogue: 0,0:27:06.76,0:27:12.43,Default,,0000,0000,0000,,before. The end effect wasn't big, but it\Nwas there and it was significant. We also, of
Dialogue: 0,0:27:12.43,0:27:17.69,Default,,0000,0000,0000,,course, considered the alternative, and the\Nmagic word is here it's.. it's Granger
Dialogue: 0,0:27:17.69,0:27:24.17,Default,,0000,0000,0000,,causality, so you can actually calculate,\Nand reliably calculate, the temporal
Dialogue: 0,0:27:24.17,0:27:34.56,Default,,0000,0000,0000,,succession, means that one follows the\Nother. And so all of a sudden, it becomes
Dialogue: 0,0:27:34.56,0:27:40.81,Default,,0000,0000,0000,,a bit difficult to point the fingers at\Nthe media. Because, if the media just
Dialogue: 0,0:27:40.81,0:27:46.46,Default,,0000,0000,0000,,reacts to an interest, it operates like a\Nbusiness. If we like it or not. There's
Dialogue: 0,0:27:46.46,0:27:50.24,Default,,0000,0000,0000,,the normative idea of the media, especially\Nin a country that is rich in high-quality
Dialogue: 0,0:27:50.24,0:27:57.23,Default,,0000,0000,0000,,publications as is Germany, that the media\Nis a public good, that educates people and
Dialogue: 0,0:27:57.23,0:28:01.98,Default,,0000,0000,0000,,brings out the best in them, in\Nchallenging them, and persuading them of
Dialogue: 0,0:28:01.98,0:28:05.76,Default,,0000,0000,0000,,the best side of the argument. But at the\Nend of the day, in your online worlds, it
Dialogue: 0,0:28:05.76,0:28:09.25,Default,,0000,0000,0000,,is a business with a measurable outcome.\NYou have clicks, you have trackers, we
Dialogue: 0,0:28:09.25,0:28:14.73,Default,,0000,0000,0000,,have ad durations that you can measure.\NAnd so you can see, which articles are
Dialogue: 0,0:28:14.73,0:28:22.77,Default,,0000,0000,0000,,favored and which articles people\Nlast the longest on. And we're not saying
Dialogue: 0,0:28:22.77,0:28:26.00,Default,,0000,0000,0000,,- there's important distinction to make\Nhere - we're not saying, that there's a
Dialogue: 0,0:28:26.00,0:28:30.73,Default,,0000,0000,0000,,direct causal link between people googling\Nrefugees and the media directly
Dialogue: 0,0:28:30.73,0:28:36.65,Default,,0000,0000,0000,,reacts to that prompt, because there's\Nsome search engine optimization guy or
Dialogue: 0,0:28:36.65,0:28:43.31,Default,,0000,0000,0000,,girl.. every media publishing house,\Nthat monitors what people are interested
Dialogue: 0,0:28:43.31,0:28:48.01,Default,,0000,0000,0000,,in. We're saying, that there's an\Nintermediate step there. It's not a
Dialogue: 0,0:28:48.01,0:28:55.03,Default,,0000,0000,0000,,direct cause, it's just a sort of delay,\Nthat is in there, that allows for other
Dialogue: 0,0:28:55.03,0:29:00.85,Default,,0000,0000,0000,,mechanisms to get in. So we're\Nwondering: What about the consumer focusing
Dialogue: 0,0:29:00.85,0:29:08.03,Default,,0000,0000,0000,,on the demand side? And in 2017, there's a\Nfew things that you could actually look at
Dialogue: 0,0:29:08.03,0:29:18.27,Default,,0000,0000,0000,,to gauge what the demand-side demands, and\Nwe decided to focus on Twitter. Because,
Dialogue: 0,0:29:18.27,0:29:21.03,Default,,0000,0000,0000,,without actually knowing this, when we\Nfirst started out with collecting all this
Dialogue: 0,0:29:21.03,0:29:27.22,Default,,0000,0000,0000,,data, we decided to set up, yeah, to set up\Na Twitter scraper. And that way, between
Dialogue: 0,0:29:27.22,0:29:32.64,Default,,0000,0000,0000,,September 1st and September 24th, we\Ncollected 4.5 million tweets, that
Dialogue: 0,0:29:32.64,0:29:40.11,Default,,0000,0000,0000,,contained keywords.. that contain any one of\Na list of keywords, that had
Dialogue: 0,0:29:40.11,0:29:50.18,Default,,0000,0000,0000,,politic connotation. So looking at this\Nbody of data, we can extract things like
Dialogue: 0,0:29:50.18,0:29:58.46,Default,,0000,0000,0000,,the top 200 most used Hashtags. And if we\Ndo that and we, we count the tweets that
Dialogue: 0,0:29:58.46,0:30:04.54,Default,,0000,0000,0000,,contains one of the top 200 Hashtags and\Nwe pay special attention, to which one of
Dialogue: 0,0:30:04.54,0:30:12.76,Default,,0000,0000,0000,,these Hashtags are decidedly pro AFD, we\Nget to a number, that 30.9 % of the tweets
Dialogue: 0,0:30:12.76,0:30:17.55,Default,,0000,0000,0000,,that contained any of those top 200\NHashtags, actually contain one that is in
Dialogue: 0,0:30:17.55,0:30:24.09,Default,,0000,0000,0000,,favor of the AFD. Whereas if we count the\Ndecidedly no AFD, the anti AFD, no AFD in
Dialogue: 0,0:30:24.09,0:30:30.76,Default,,0000,0000,0000,,all ways of spelling and capitalization\Nand so forth, that's only 1.2 %.
Dialogue: 0,0:30:30.76,0:30:38.78,Default,,0000,0000,0000,,And here it becomes a bit\Nticklish. So, in order to sort of give a
Dialogue: 0,0:30:38.78,0:30:46.02,Default,,0000,0000,0000,,better idea, of what role Twitter might\Nhave played in our little, in our little
Dialogue: 0,0:30:46.02,0:30:51.05,Default,,0000,0000,0000,,relationship here, between the demand side\Nand the supply side, the supply side
Dialogue: 0,0:30:51.05,0:30:59.28,Default,,0000,0000,0000,,supplying the news, we have a\Nbeautiful network graph. So this is a
Dialogue: 0,0:30:59.28,0:31:08.14,Default,,0000,0000,0000,,retweeting network: this is we extract all\Nthe mentions of a.. of an actor. Each dot is
Dialogue: 0,0:31:08.14,0:31:13.61,Default,,0000,0000,0000,,a Twitter user, each line is 5 or more\Nretweets. Retweets, we're aware of that,
Dialogue: 0,0:31:13.61,0:31:17.22,Default,,0000,0000,0000,,Retweets don't automatically mean\Nendorsement - you might retweet something
Dialogue: 0,0:31:17.22,0:31:24.55,Default,,0000,0000,0000,,that is outlandish and crazy. But for the\Nsake of visualizing, what the weights are
Dialogue: 0,0:31:24.55,0:31:29.59,Default,,0000,0000,0000,,on Twitter, we're treating them as the\Nsame. And anyone, who ever has worked
Dialogue: 0,0:31:29.59,0:31:31.97,Default,,0000,0000,0000,,with network graphs of that size, that\Ntake a long time
Dialogue: 0,0:31:31.97,0:31:36.84,Default,,0000,0000,0000,,to generate, and it's kind of\Ntough to to label them, so I'm very proud,
Dialogue: 0,0:31:36.84,0:31:43.75,Default,,0000,0000,0000,,that I was able to do so. If we look at\Nthis Island down there, that blob, that
Dialogue: 0,0:31:43.75,0:31:53.64,Default,,0000,0000,0000,,blue blob - those are accounts, that\Ncluster around AFD accounts. The coloring
Dialogue: 0,0:31:53.64,0:31:58.49,Default,,0000,0000,0000,,here was done by a walk trap\Nalgorithm, I just adjusted the colors that
Dialogue: 0,0:31:58.49,0:32:03.14,Default,,0000,0000,0000,,that algorithm used to actually match the\Ncolors in the, in the German party
Dialogue: 0,0:32:03.14,0:32:11.00,Default,,0000,0000,0000,,landscape. And so we do have a hefty\Ncontinent at the bottom right, that
Dialogue: 0,0:32:11.00,0:32:16.73,Default,,0000,0000,0000,,connects all kinds of people to the\NAFD. There, if you look at the little
Dialogue: 0,0:32:16.73,0:32:24.01,Default,,0000,0000,0000,,appendix below here, that is colored in\Nbrown, that is mainly organized around a
Dialogue: 0,0:32:24.01,0:32:31.84,Default,,0000,0000,0000,,movement called Reconquista Europe, which\Nis an even further right wing, right wing
Dialogue: 0,0:32:31.84,0:32:40.84,Default,,0000,0000,0000,,movement that is sort of, like, directly\Ntacked to this island of the AFD, and the
Dialogue: 0,0:32:40.84,0:32:47.76,Default,,0000,0000,0000,,the connecting node is Björn Höcke, which\Nis quite interesting. So we have the AFD down
Dialogue: 0,0:32:47.76,0:32:52.66,Default,,0000,0000,0000,,there, we have the other parties up there, the\Nrainbow, that is the pluralistic political
Dialogue: 0,0:32:52.66,0:32:59.75,Default,,0000,0000,0000,,landscape, we have those.. those two extreme\Npoints there, at the at the super top right
Dialogue: 0,0:32:59.75,0:33:08.11,Default,,0000,0000,0000,,and there at the bottom left - that is..\Nthose are very extremely.. extreme Twitter
Dialogue: 0,0:33:08.11,0:33:13.59,Default,,0000,0000,0000,,user parties. That's the ÖDP and the\NFreien Wähler, so they don't.. they don't seem to
Dialogue: 0,0:33:13.59,0:33:19.37,Default,,0000,0000,0000,,engage with the nodes, that are in the\Ncenter here. But what's also valiable to
Dialogue: 0,0:33:19.37,0:33:25.06,Default,,0000,0000,0000,,note, is that for the other parties, for the\Nestablished parties, starting from the left
Dialogue: 0,0:33:25.06,0:33:30.48,Default,,0000,0000,0000,,and orange - the Pirate Party and then red\Nthe Social Democrats, Purple is Die Linke,
Dialogue: 0,0:33:30.48,0:33:37.16,Default,,0000,0000,0000,,green die Grünen, yellow FDP and black the\NConservative Party CDU/CSU. All of these
Dialogue: 0,0:33:37.16,0:33:44.03,Default,,0000,0000,0000,,parties have a central node - a central, a\Ncentral account, around which a lot of
Dialogue: 0,0:33:44.03,0:33:50.39,Default,,0000,0000,0000,,other users are fanned out. So there's.. for\Neach party, there's a smaller number, or a
Dialogue: 0,0:33:50.39,0:33:55.59,Default,,0000,0000,0000,,relatively small number, of accounts, that\Nare highly favored in how often they are
Dialogue: 0,0:33:55.59,0:34:04.26,Default,,0000,0000,0000,,retweeted. AFD doesn't have that. Even.. so,\Nthis is of course a projection of
Dialogue: 0,0:34:04.26,0:34:07.31,Default,,0000,0000,0000,,something that's 3D in a\N2D place, so there might be
Dialogue: 0,0:34:07.31,0:34:11.24,Default,,0000,0000,0000,,some skewing going on here, in terms of how\Nit shows on our screen, but even turning it
Dialogue: 0,0:34:11.24,0:34:18.79,Default,,0000,0000,0000,,and trying to identify which party is at\Nthe center, wasn't, wasn't really possible.
Dialogue: 0,0:34:18.79,0:34:26.01,Default,,0000,0000,0000,,So the internal rifts, and the internal\Npower struggles - they do show in how
Dialogue: 0,0:34:26.01,0:34:32.36,Default,,0000,0000,0000,,members of the party are retweeted. Also\Ninteresting to note is which nodes, which
Dialogue: 0,0:34:32.36,0:34:38.77,Default,,0000,0000,0000,,users are connecting these two continents,\Nso to speak. One is, that blue dot, is
Dialogue: 0,0:34:38.77,0:34:41.86,Default,,0000,0000,0000,,wahlrecht.de, a polling\Naggregator, of course, everyone is
Dialogue: 0,0:34:41.86,0:34:45.84,Default,,0000,0000,0000,,interested in getting their polling\Nnumbers out. And there's.. that's tough to
Dialogue: 0,0:34:45.84,0:34:53.25,Default,,0000,0000,0000,,see here, but there's a beige user in\Nthe middle therem which is welt.de, so
Dialogue: 0,0:34:53.25,0:34:58.02,Default,,0000,0000,0000,,one of the media.. one of the media\Npublications, that we actually collected
Dialogue: 0,0:34:58.02,0:35:02.97,Default,,0000,0000,0000,,data on and surveyed. Another thing that\Nis.. I'm just gonna mention here briefly, is
Dialogue: 0,0:35:02.97,0:35:13.83,Default,,0000,0000,0000,,the.. that light pink colored insert between\Nthe greens and the central gray beige dot
Dialogue: 0,0:35:13.83,0:35:17.51,Default,,0000,0000,0000,,- those are Jan Böhmermann, die Heute-show\Nund Extra 3.
Dialogue: 0,0:35:17.51,0:35:26.46,Default,,0000,0000,0000,,{\i1}laughing, clapping{\i0}\NYeah, so there's a.. the dynamics are
Dialogue: 0,0:35:26.46,0:35:32.75,Default,,0000,0000,0000,,clear, that we have this party that is\Npretty well organized on social media,
Dialogue: 0,0:35:32.75,0:35:40.28,Default,,0000,0000,0000,,and thus is able to dominate a media\Nagenda, that is based on algorithms basically.
Dialogue: 0,0:35:40.28,0:35:46.98,Default,,0000,0000,0000,,If you think about how, how the logic of\Ninformation dissemination works on
Dialogue: 0,0:35:46.98,0:35:54.70,Default,,0000,0000,0000,,Twitter: with trending Hashtags. If you\Nhave a party that is as - well, I don't want
Dialogue: 0,0:35:54.70,0:35:58.70,Default,,0000,0000,0000,,to say organized - but as tightly clustered\Naround itself.. within itself as the AFD
Dialogue: 0,0:35:58.70,0:36:05.43,Default,,0000,0000,0000,,shows up here, there's a good chance, that\Nthat will influence, what all of us get to
Dialogue: 0,0:36:05.43,0:36:11.80,Default,,0000,0000,0000,,see, when we check out the Twitter homepage.\NNow I know, that probably a good chunk of
Dialogue: 0,0:36:11.80,0:36:18.08,Default,,0000,0000,0000,,you have burning questions in their mind,\Nand I'm gonna going to want to know - so
Dialogue: 0,0:36:18.08,0:36:25.73,Default,,0000,0000,0000,,how many of these of these bright blue bar\Nblobs are bots? Are Twitter bots. We tried to
Dialogue: 0,0:36:25.73,0:36:31.51,Default,,0000,0000,0000,,find that out, using a tool called the \NBotometer, which is something that has an
Dialogue: 0,0:36:31.51,0:36:35.74,Default,,0000,0000,0000,,API available online, where you can submit,\Nit's a project from a research team
Dialogue: 0,0:36:35.74,0:36:40.51,Default,,0000,0000,0000,,in Indiana, where you can submit the name\Nof a Twitter user and then it gives you a..
Dialogue: 0,0:36:40.51,0:36:46.32,Default,,0000,0000,0000,,it rends lots of lots of analyses and\Nanalysis lots of things
Dialogue: 0,0:36:46.32,0:36:51.61,Default,,0000,0000,0000,,about this user: the frequency of tweets,\Nthe time at which it tweets, who is.. who
Dialogue: 0,0:36:51.61,0:36:56.17,Default,,0000,0000,0000,,is it following, who is it followed by, who\Nis it talking to, that kind of stuff. But
Dialogue: 0,0:36:56.17,0:37:03.64,Default,,0000,0000,0000,,when I tried to submit that, I broke their\NAPI. And so, if they're happen to watch I
Dialogue: 0,0:37:03.64,0:37:09.90,Default,,0000,0000,0000,,apologize, that was me. So it wasn't be\Nable to do so in time, but there's a bunch of
Dialogue: 0,0:37:09.90,0:37:13.98,Default,,0000,0000,0000,,talks tomorrow, that talk about exactly..\Nabout that thing, so I'm happy to have
Dialogue: 0,0:37:13.98,0:37:20.21,Default,,0000,0000,0000,,this sort of as a lead-in for the day\Ntomorrow. So what can we.. go, what can we
Dialogue: 0,0:37:20.21,0:37:27.09,Default,,0000,0000,0000,,take from this? The Bundestagswahl 2017 was a\Nperfect storm for a far-right party like
Dialogue: 0,0:37:27.09,0:37:31.57,Default,,0000,0000,0000,,the AFD. You had a high issue salience of\Nthe topic that is at the center of its
Dialogue: 0,0:37:31.57,0:37:42.37,Default,,0000,0000,0000,,agenda, and you have a sort of unregulated\NWild West of social media. We'll see
Dialogue: 0,0:37:42.37,0:37:46.93,Default,,0000,0000,0000,,how that changes with recent law\Nchanges come into effect, where all of a
Dialogue: 0,0:37:46.93,0:37:52.55,Default,,0000,0000,0000,,sudden the platform itself has some\Nliability, to which kind of messages are
Dialogue: 0,0:37:52.55,0:37:57.31,Default,,0000,0000,0000,,spread. But if that's effective for\NTwitter, is a whole other bag of
Dialogue: 0,0:37:57.31,0:38:03.86,Default,,0000,0000,0000,,worms. So in that sense, that's why I was..\Nwhat I was sorting.. hinting at: in this
Dialogue: 0,0:38:03.86,0:38:08.64,Default,,0000,0000,0000,,issue environment, we have people be\Ninterested in the topic that is central
Dialogue: 0,0:38:08.64,0:38:16.76,Default,,0000,0000,0000,,for the party like the AFDs. The media\Nbehaved like pretty surprisingly..
Dialogue: 0,0:38:16.76,0:38:22.99,Default,,0000,0000,0000,,surprisingly predictable?\NAnd did not.. at least for the.. for the
Dialogue: 0,0:38:22.99,0:38:30.65,Default,,0000,0000,0000,,topics, or for the publications, that we\Ncovered, it did so. And for the context
Dialogue: 0,0:38:30.65,0:38:35.66,Default,,0000,0000,0000,,that we're arguing herein, that the AFD\Nonly get like 20 % of the share
Dialogue: 0,0:38:35.66,0:38:40.00,Default,,0000,0000,0000,,towards the end of the campaign, is\Nsomething that is a little bit surprising.
Dialogue: 0,0:38:40.00,0:38:46.82,Default,,0000,0000,0000,,And that also leads into a different\Nquestion of what does this "Oh it's the
Dialogue: 0,0:38:46.82,0:38:54.75,Default,,0000,0000,0000,,journalists fault!" actually mean? What\Ndoes it really mean? This sort of is based
Dialogue: 0,0:38:54.75,0:39:00.66,Default,,0000,0000,0000,,on this normative expectation of the media\Nbeing an impartial.. an impartial
Dialogue: 0,0:39:00.66,0:39:05.92,Default,,0000,0000,0000,,deliverer of information and if you think\Nabout what else is going on on the
Dialogue: 0,0:39:05.92,0:39:11.54,Default,,0000,0000,0000,,internet, with alternative media and an\Nalternative news sphere establishing
Dialogue: 0,0:39:11.54,0:39:18.39,Default,,0000,0000,0000,,itself with news blogs like, well I\Ndon't wanna.. I don't want to call any
Dialogue: 0,0:39:18.39,0:39:25.73,Default,,0000,0000,0000,,names, because.. and so, there's a sort of\Nscene of far-right fringe blogs in Germany
Dialogue: 0,0:39:25.73,0:39:29.60,Default,,0000,0000,0000,,that we also collected.\NAnd so we're.. further down the line,
Dialogue: 0,0:39:29.60,0:39:34.33,Default,,0000,0000,0000,,we're going to look at what the topics\Nwere, that were covered in that and how
Dialogue: 0,0:39:34.33,0:39:39.48,Default,,0000,0000,0000,,that connected to influencing public\Nopinion in Germany, but having said this,
Dialogue: 0,0:39:39.48,0:39:44.92,Default,,0000,0000,0000,,with these alternative ways of getting\Nyour news, information being available, if
Dialogue: 0,0:39:44.92,0:39:49.93,Default,,0000,0000,0000,,you have the press, if you have the\Nmainstream press, not covering a party like
Dialogue: 0,0:39:49.93,0:39:56.03,Default,,0000,0000,0000,,the AFD to a certain extent, you only give\Nthe fodder to those cries of
Dialogue: 0,0:39:56.03,0:40:07.26,Default,,0000,0000,0000,,"Lügenpresse", mendacious press, in\Nmembers of the population, that are sort of
Dialogue: 0,0:40:07.26,0:40:10.67,Default,,0000,0000,0000,,at the risk of being lost as audience\Nmembers.
Dialogue: 0,0:40:10.67,0:40:14.95,Default,,0000,0000,0000,,So, it's kind of difficult to call the..\Nto call the shots here and actually point
Dialogue: 0,0:40:14.95,0:40:18.24,Default,,0000,0000,0000,,the fingers at the media, because they\Ndelivered on informing on an interest that
Dialogue: 0,0:40:18.24,0:40:30.05,Default,,0000,0000,0000,,existed in the in the population, before\Nthey reported on something like the AFD.
Dialogue: 0,0:40:30.05,0:40:33.52,Default,,0000,0000,0000,,And with this, I want to leave it at that.\NI thank you very much for your attention
Dialogue: 0,0:40:33.52,0:40:38.50,Default,,0000,0000,0000,,and I'm highly, highly eager to hear\Nquestions and prompts and ideas, how we
Dialogue: 0,0:40:38.50,0:40:40.75,Default,,0000,0000,0000,,could pursue this further.
Dialogue: 0,0:40:40.75,0:40:47.73,Default,,0000,0000,0000,,{\i1}applause{\i0}
Dialogue: 0,0:40:47.73,0:40:55.52,Default,,0000,0000,0000,,Herald: Vielen Dank, thank you very much.\NQuestions? [unintelligible] any questions?
Dialogue: 0,0:40:55.52,0:41:00.48,Default,,0000,0000,0000,,Feel free to attend the microphones.\NEven the microphone I don't
Dialogue: 0,0:41:00.48,0:41:05.75,Default,,0000,0000,0000,,see behind the cameras. Let's start with\Nnumber two. {\i1}laughs{\i0}
Dialogue: 0,0:41:05.75,0:41:08.72,Default,,0000,0000,0000,,Mic 2: [unintelligible] some sound. \NThank you. Thank you very much for your
Dialogue: 0,0:41:08.72,0:41:15.76,Default,,0000,0000,0000,,amazing work. I've got only one question.\NDo you plan on releasing those
Dialogue: 0,0:41:15.76,0:41:21.48,Default,,0000,0000,0000,,collected data and on what license?\NBeyer: That's a question that we.. that
Dialogue: 0,0:41:21.48,0:41:27.92,Default,,0000,0000,0000,,we asked ourselves, too. We would love to\Ncollect the data and ultimately it will
Dialogue: 0,0:41:27.92,0:41:31.04,Default,,0000,0000,0000,,happen, but we have to make sure, that we\Nactually have the right to do so, with the
Dialogue: 0,0:41:31.04,0:41:35.77,Default,,0000,0000,0000,,way we collected it. But we're definitely\Nlooking into that.
Dialogue: 0,0:41:35.77,0:41:44.19,Default,,0000,0000,0000,,Herald: OK, number 5. Yeah, you.\NMic 5: OK. Hello. Is this working? Yeah. It's
Dialogue: 0,0:41:44.19,0:41:48.17,Default,,0000,0000,0000,,tempting - I'm from the Netherlands -\Nto compare these experiences with the AFD
Dialogue: 0,0:41:48.17,0:41:52.27,Default,,0000,0000,0000,,with the experience in the Netherlands.\NYou know, we had Wilders, we had Verdonk
Dialogue: 0,0:41:52.27,0:41:57.74,Default,,0000,0000,0000,,we had Fortuyn, now we have Baudet and it\Nseems that there is a major difference
Dialogue: 0,0:41:57.74,0:42:02.36,Default,,0000,0000,0000,,between.. with the AFD, because presently,\NI have frankly, I don't know the name of
Dialogue: 0,0:42:02.36,0:42:07.55,Default,,0000,0000,0000,,the leader of the AFD, it used to be\NFrauke Petry and now, I don't know.
Dialogue: 0,0:42:07.55,0:42:15.76,Default,,0000,0000,0000,,But in the Netherlands, the leaders of the..\Nthose populist right-wing parties, they
Dialogue: 0,0:42:15.76,0:42:20.99,Default,,0000,0000,0000,,were.. they were very good in manipulating\Nthe media. They were sending out messages
Dialogue: 0,0:42:20.99,0:42:27.54,Default,,0000,0000,0000,,sustaining a "Köder", in Germany, what's the\Nword? Like, um, provocating, sending
Dialogue: 0,0:42:27.54,0:42:32.31,Default,,0000,0000,0000,,out provocations and that attracted \Nattention of the media. So, there are
Dialogue: 0,0:42:32.31,0:42:37.07,Default,,0000,0000,0000,,people saying that you shouldn't react on\Nall provocations, but anyway they were
Dialogue: 0,0:42:37.07,0:42:43.57,Default,,0000,0000,0000,,geared to draw attention and I wonder,\Nwhether AFD has been to the same extent
Dialogue: 0,0:42:43.57,0:42:52.52,Default,,0000,0000,0000,,active in the field of drawing attention,\Npurposely using even agencies that are
Dialogue: 0,0:42:52.52,0:42:58.23,Default,,0000,0000,0000,,specialized in advertising.\NBeyer: Great question. There is this idea,
Dialogue: 0,0:42:58.23,0:43:05.10,Default,,0000,0000,0000,,that the AFD was very skillful at sort of\Ninscenating scandal and purposely doing
Dialogue: 0,0:43:05.10,0:43:11.51,Default,,0000,0000,0000,,things on a public stage that would draw\Nattention to them. For example, this..
Dialogue: 0,0:43:11.51,0:43:17.25,Default,,0000,0000,0000,,yeah, I say it again.. this\Nexpression by Alexander Gauland, to dispose
Dialogue: 0,0:43:17.25,0:43:24.03,Default,,0000,0000,0000,,of a German politician, or the other\Nleading candidate Alice Weidel leaving a
Dialogue: 0,0:43:24.03,0:43:32.31,Default,,0000,0000,0000,,talk show, while it was being broadcast. So\Nthere's.. there definitely is this
Dialogue: 0,0:43:32.31,0:43:37.35,Default,,0000,0000,0000,,element of the.. of actually taking a\Nscandal and using it for your own, for
Dialogue: 0,0:43:37.35,0:43:44.98,Default,,0000,0000,0000,,pushing your own agenda, whereas if they\Nused ad agencies for their media campaign,
Dialogue: 0,0:43:44.98,0:43:53.27,Default,,0000,0000,0000,,they did, their campaigning was highly\Nprofessionalized, in terms of what their
Dialogue: 0,0:43:53.27,0:43:57.37,Default,,0000,0000,0000,,posters were and how their campaign ads\Nwere worked.
Dialogue: 0,0:43:57.37,0:44:05.40,Default,,0000,0000,0000,,And they did work with a company, that also\Nwas involved with Donald Trump's campaign.
Dialogue: 0,0:44:05.40,0:44:12.68,Default,,0000,0000,0000,,But in terms of.. sort of new media or\Nlike online media – it's not that new
Dialogue: 0,0:44:12.68,0:44:19.35,Default,,0000,0000,0000,,anymore – and in terms of what they did on\Nonline media, I.. I only have an
Dialogue: 0,0:44:19.35,0:44:28.57,Default,,0000,0000,0000,,anecdotal sense, if they use something like\Nbots, which is also a way of buying,
Dialogue: 0,0:44:28.57,0:44:34.76,Default,,0000,0000,0000,,buying attention. I can.. I can sort\Nof tell you about one specific case, where
Dialogue: 0,0:44:34.76,0:44:42.22,Default,,0000,0000,0000,,we investigated, which Twitter users were\Nthe most active in tweeting on the AFD on
Dialogue: 0,0:44:42.22,0:44:47.37,Default,,0000,0000,0000,,German Twitter – tomorrow's a talk about a\NTwitter user called Ballerina, which is
Dialogue: 0,0:44:47.37,0:44:50.46,Default,,0000,0000,0000,,a name that has been out there which..\Nthere's great education, that that is
Dialogue: 0,0:44:50.46,0:44:54.64,Default,,0000,0000,0000,,definitely a bot, that has been planted\Nand has been controlled by someone else or
Dialogue: 0,0:44:54.64,0:45:06.00,Default,,0000,0000,0000,,by sort of.. by any group of\Nactors that is not actually a ballerina.
Dialogue: 0,0:45:06.00,0:45:14.21,Default,,0000,0000,0000,,What we found was a Twitter user called\NTeletubbies007, that tweeted in those three
Dialogue: 0,0:45:14.21,0:45:22.90,Default,,0000,0000,0000,,weeks, that we surveyed, 6.500 times and\Nmostly just retweeted, retweeted calls to
Dialogue: 0,0:45:22.90,0:45:29.99,Default,,0000,0000,0000,,go and cast your ballot, that were all put\Nout by the central AFD accounts. And it
Dialogue: 0,0:45:29.99,0:45:33.71,Default,,0000,0000,0000,,didn't have a lot of followers, like\Nsomething 500 or so, but it just kept
Dialogue: 0,0:45:33.71,0:45:39.85,Default,,0000,0000,0000,,retweeting over and over and over and over.\NAnd when we actually wanted to check out
Dialogue: 0,0:45:39.85,0:45:46.36,Default,,0000,0000,0000,,the page of that bot, it was deleted,\Nthe user was deleted.
Dialogue: 0,0:45:46.36,0:45:51.14,Default,,0000,0000,0000,,So there's, to answer your question, um,\Nthis high.. this degree of
Dialogue: 0,0:45:51.14,0:46:00.04,Default,,0000,0000,0000,,personalization that the Partij voor\Nde Vrijheid has in the Netherlands is not
Dialogue: 0,0:46:00.04,0:46:03.18,Default,,0000,0000,0000,,as extreme for the AFD in Germany, because\Nthere's more leading candidates and
Dialogue: 0,0:46:03.18,0:46:09.30,Default,,0000,0000,0000,,there's internal rifts like Geerd Wilders\Nis basically his own party. That's not the
Dialogue: 0,0:46:09.30,0:46:15.24,Default,,0000,0000,0000,,same. But the strategy to use scandal and\Nto use something that is outrageous and
Dialogue: 0,0:46:15.24,0:46:18.97,Default,,0000,0000,0000,,push the boundaries a little bit more,\Nthen jump back and say "Oh no, we did not
Dialogue: 0,0:46:18.97,0:46:24.26,Default,,0000,0000,0000,,mean that at all in this way", that is\Nthe exact same spot on strategy they used.
Dialogue: 0,0:46:24.26,0:46:27.63,Default,,0000,0000,0000,,Mic 5: Perhaps I should add that Wilders\Nmade it like..
Dialogue: 0,0:46:27.63,0:46:31.53,Default,,0000,0000,0000,,Herald: Excuse me, many people queuing.\NMic 5: Okay. Then I'll stop.
Dialogue: 0,0:46:31.53,0:46:35.02,Default,,0000,0000,0000,,Herald: Okay, thank you. We have questions\Nfrom the internet, then.
Dialogue: 0,0:46:35.02,0:46:40.66,Default,,0000,0000,0000,,Signal Angel: Yes. (?) is asking: "Why\Ndid you come to the conclusion that this
Dialogue: 0,0:46:40.66,0:46:44.62,Default,,0000,0000,0000,,was a special election, while the last\Nelection in Austria has exactly the same
Dialogue: 0,0:46:44.62,0:46:49.35,Default,,0000,0000,0000,,issues? Don't you see this as some sort of\Nan global effect?"
Dialogue: 0,0:46:49.35,0:46:57.35,Default,,0000,0000,0000,,Beyer: That's true, a Syrian civil war\Nthat pushes people to flee from,
Dialogue: 0,0:46:57.35,0:47:00.46,Default,,0000,0000,0000,,from war and save their livelihood, is\Nsomething that is not only felt in
Dialogue: 0,0:47:00.46,0:47:05.05,Default,,0000,0000,0000,,Germany, but for the context of Germany,\Nit's a special election. That's.. this sort
Dialogue: 0,0:47:05.05,0:47:11.33,Default,,0000,0000,0000,,of situations never.. has never occurred in\Nthis way before. But absolutely, each
Dialogue: 0,0:47:11.33,0:47:20.74,Default,,0000,0000,0000,,election in Europe basically since 2015\Nwas a special election in that sense. But
Dialogue: 0,0:47:20.74,0:47:25.47,Default,,0000,0000,0000,,not in terms of the outcomes, in a way,\Nthat.. because far-right parties in other
Dialogue: 0,0:47:25.47,0:47:30.50,Default,,0000,0000,0000,,European countries already had, had their\Nfoot in the door and especially in Austria,
Dialogue: 0,0:47:30.50,0:47:35.55,Default,,0000,0000,0000,,where.. with the FPÖ were pretty well\Nestablished with previously having been
Dialogue: 0,0:47:35.55,0:47:39.80,Default,,0000,0000,0000,,part of a government. And now\Nbeing part of the government again. But
Dialogue: 0,0:47:39.80,0:47:45.80,Default,,0000,0000,0000,,for Germany, in what the issues were, that\Nwere top of people's minds: that's the
Dialogue: 0,0:47:45.80,0:47:51.37,Default,,0000,0000,0000,,special case that I meant.\NHerald: OK, microphone number 3, please.
Dialogue: 0,0:47:51.37,0:47:58.12,Default,,0000,0000,0000,,Mic 3: Thank you, first I really\Nappreciate the sincerity and transparency
Dialogue: 0,0:47:58.12,0:48:03.01,Default,,0000,0000,0000,,of your talk, thank you very much, we need\Nmore of this in such circumstances and
Dialogue: 0,0:48:03.01,0:48:10.66,Default,,0000,0000,0000,,maybe less polemics sometimes. There's\Njust a little trifle in your method, where
Dialogue: 0,0:48:10.66,0:48:18.19,Default,,0000,0000,0000,,I was wondering: how did you filter the\N"Linke" and "Grüne" stuff. Did you..
Dialogue: 0,0:48:18.19,0:48:23.76,Default,,0000,0000,0000,,yeah, how exactly did you do it? Did you\Nmaybe count all the mentions of "Grün"
Dialogue: 0,0:48:23.76,0:48:29.84,Default,,0000,0000,0000,,with a capital and non-capital "G", and\N"Linke" with a capital L and non-capital
Dialogue: 0,0:48:29.84,0:48:34.89,Default,,0000,0000,0000,,and then filter it out further? Or did you\Ndo it the other way around? I know, that
Dialogue: 0,0:48:34.89,0:48:41.31,Default,,0000,0000,0000,,you focused specifically on the AFD stuff,\Nand maybe you were focused on representing
Dialogue: 0,0:48:41.31,0:48:46.57,Default,,0000,0000,0000,,all the parties that might be relevant.\NBut I would still be interested in that
Dialogue: 0,0:48:46.57,0:48:50.85,Default,,0000,0000,0000,,part, thanks.\NBeyer: That's a great question. The thing
Dialogue: 0,0:48:50.85,0:48:55.45,Default,,0000,0000,0000,,is, that we used.. when we actually put all\Nthat.. when after we collected text, before
Dialogue: 0,0:48:55.45,0:49:01.51,Default,,0000,0000,0000,,we put it through the unloading\Nmethods, we put it all into lowercase.
Dialogue: 0,0:49:01.51,0:49:09.58,Default,,0000,0000,0000,,Just so we could have a consistent way of\Nanalyzing. And with capitalization, it's
Dialogue: 0,0:49:09.58,0:49:16.29,Default,,0000,0000,0000,,kind of.. sometimes it just trips up the\Nway to treat this. And that's why you
Dialogue: 0,0:49:16.29,0:49:21.37,Default,,0000,0000,0000,,ran into these issues with "Linke und\NGrüne" where we had to resort to only
Dialogue: 0,0:49:21.37,0:49:25.51,Default,,0000,0000,0000,,taking basically the candidates names and\Nthen also "Linke Partei" and "Grüne
Dialogue: 0,0:49:25.51,0:49:35.75,Default,,0000,0000,0000,,Partei" and a few conjugations, so "der Linken\NPartei", "den".. right, like grammatically..
Dialogue: 0,0:49:35.75,0:49:40.50,Default,,0000,0000,0000,,the cases, we only, like, we conjugated\Nthem through. Yeah, but we.. since our focus
Dialogue: 0,0:49:40.50,0:49:43.91,Default,,0000,0000,0000,,was on the AfD, we weren't especially\Nconcerned with that, which is
Dialogue: 0,0:49:43.91,0:49:48.35,Default,,0000,0000,0000,,unfortunate, I admit that, but for the\Npurpose of this talk, we decided to just
Dialogue: 0,0:49:48.35,0:49:53.50,Default,,0000,0000,0000,,use this workaround.\NMic 3: OK, thanks.
Dialogue: 0,0:49:53.50,0:49:56.26,Default,,0000,0000,0000,,Herald: OK. Microphone six,\Nplease.
Dialogue: 0,0:49:56.26,0:49:59.13,Default,,0000,0000,0000,,Mic 6: Hello, thanks for your interesting\Npresentation.
Dialogue: 0,0:49:59.13,0:50:03.48,Default,,0000,0000,0000,,I'm wondering if you and your team..\Nso, you'd.. you looked at mentions
Dialogue: 0,0:50:03.48,0:50:07.76,Default,,0000,0000,0000,,of the different parties, but I'm wondering\Nif you looked at the content of the articles
Dialogue: 0,0:50:07.76,0:50:12.65,Default,,0000,0000,0000,,and how they talked about it, if they were\Ntalked about positively or negatively.
Dialogue: 0,0:50:12.65,0:50:15.11,Default,,0000,0000,0000,,Beyer: Thank you very much, that's a great
Dialogue: 0,0:50:15.11,0:50:19.07,Default,,0000,0000,0000,,question, that we actually did consider.\NAnd I'll answer this question with a
Dialogue: 0,0:50:19.07,0:50:23.52,Default,,0000,0000,0000,,counter question, as social scientists\Nlike to do. Anyone in this room use Amazon
Dialogue: 0,0:50:23.52,0:50:28.59,Default,,0000,0000,0000,,Mechanical Turk and works on hits to earn\Na few cents here and there? No? OK, so I
Dialogue: 0,0:50:28.59,0:50:37.58,Default,,0000,0000,0000,,can speak freely. There's a.. there's a\Nmethod that uses cheap labor on Amazon
Dialogue: 0,0:50:37.58,0:50:42.92,Default,,0000,0000,0000,,Mechanical Turk and presents each worker\Nwith two sentences, out of which they have
Dialogue: 0,0:50:42.92,0:50:50.00,Default,,0000,0000,0000,,to change the one that is more positive.\NAnd so we wanted to use this to train a
Dialogue: 0,0:50:50.00,0:50:55.12,Default,,0000,0000,0000,,machine learning algorithm to actually get\Na way to gauge the sentiment of positive
Dialogue: 0,0:50:55.12,0:50:58.93,Default,,0000,0000,0000,,and negative in the text that we had\Ncollected. We started that in early
Dialogue: 0,0:50:58.93,0:51:06.73,Default,,0000,0000,0000,,December and we had a, like, a workbook with\N4.000 so-called hits, 4.000 little jobs,
Dialogue: 0,0:51:06.73,0:51:16.14,Default,,0000,0000,0000,,4.000 comparisons and when this job was\Ndone, five or six days later, we sort of put
Dialogue: 0,0:51:16.14,0:51:23.49,Default,,0000,0000,0000,,that through a test and compared it with\Nour own hand-coding that we had done.
Dialogue: 0,0:51:23.49,0:51:31.56,Default,,0000,0000,0000,,And it turned out that one worker on Amazon\NMechanical Turk spent over seven hours and
Dialogue: 0,0:51:31.56,0:51:38.51,Default,,0000,0000,0000,,worked.. of those 4.000 little jobs\Nthat we had, he worked 3.980.
Dialogue: 0,0:51:38.51,0:51:45.25,Default,,0000,0000,0000,,And over 1.400 of which \Nhe did in less than two seconds.
Dialogue: 0,0:51:45.25,0:51:53.84,Default,,0000,0000,0000,,Which is unfortunate, because: a) this\Nperson.. so, this person - right, "Person?, Question
Dialogue: 0,0:51:53.84,0:51:59.87,Default,,0000,0000,0000,,Mark" - probably used a script, probably\Nused a bot or just randomly clicked. The
Dialogue: 0,0:51:59.87,0:52:06.77,Default,,0000,0000,0000,,coding didn't match up at all with what we\Ndid hand-wise ourselves and that really
Dialogue: 0,0:52:06.77,0:52:13.71,Default,,0000,0000,0000,,screwed up our approach there. If any of\Nyou plan on doing some hits in the new
Dialogue: 0,0:52:13.71,0:52:18.86,Default,,0000,0000,0000,,year for Amazon Mechanical Turk and you're\Nasked to compare two sentences that
Dialogue: 0,0:52:18.86,0:52:23.47,Default,,0000,0000,0000,,mention a political actor in Germany, you\Ncan send me an email and maybe a
Dialogue: 0,0:52:23.47,0:52:28.11,Default,,0000,0000,0000,,screenshot and tell me how much you\Nappreciate that we're paying six cents for
Dialogue: 0,0:52:28.11,0:52:34.44,Default,,0000,0000,0000,,each comparison. But that's the story, why\Nwe haven't.. we don't have any sentiment in
Dialogue: 0,0:52:34.44,0:52:40.26,Default,,0000,0000,0000,,this analysis here.\NHerald: [unintelligible]
Dialogue: 0,0:52:40.26,0:52:44.84,Default,,0000,0000,0000,,Mic ?: Hello. I'm from Denmark, so in this\Ncontext, I'm very much a ghost of\N
Dialogue: 0,0:52:44.84,0:52:47.50,Default,,0000,0000,0000,,Christmas future. \N{\i1}Beyer laughs{\i0}
Dialogue: 0,0:52:47.50,0:52:49.04,Default,,0000,0000,0000,,Mic ?: In your Twitter data,
Dialogue: 0,0:52:49.04,0:52:54.66,Default,,0000,0000,0000,,where you take Retweets as well, do you\Ndetermine what are quotes and what
Dialogue: 0,0:52:54.66,0:53:00.29,Default,,0000,0000,0000,,are direct Retweets? Because in my\Nexperience, and I work with this in Denmark
Dialogue: 0,0:53:00.29,0:53:07.36,Default,,0000,0000,0000,,and in the UK, a lot of people like to\Ndistance themselves from what the AfD and
Dialogue: 0,0:53:07.36,0:53:15.74,Default,,0000,0000,0000,,similar are saying by quoting everything\Nthey're saying and giving them the press.
Dialogue: 0,0:53:15.74,0:53:21.36,Default,,0000,0000,0000,,Beyer: That's a very good point to make.\NWe did not make any distinction
Dialogue: 0,0:53:21.36,0:53:27.02,Default,,0000,0000,0000,,between quotes and Retweets, but we did\Nfilter, based on 5 Retweets, by thinking:
Dialogue: 0,0:53:27.02,0:53:31.28,Default,,0000,0000,0000,,OK, if you occasionally feel like you\Nhave to point something out that is
Dialogue: 0,0:53:31.28,0:53:36.84,Default,,0000,0000,0000,,outrageous and ridiculous, that a person,\Na member of a party, says on Twitter, you
Dialogue: 0,0:53:36.84,0:53:41.10,Default,,0000,0000,0000,,would be inclined to do so less than a\Ncertain amount of time. We also tried it
Dialogue: 0,0:53:41.10,0:53:45.76,Default,,0000,0000,0000,,with other cut-offs. The graph basically\Nalways looked the same. But if we think
Dialogue: 0,0:53:45.76,0:53:52.70,Default,,0000,0000,0000,,about what this means for how the demand-\Nside is influenced, it doesn't matter.
Dialogue: 0,0:53:52.70,0:53:57.21,Default,,0000,0000,0000,,Basically, if you're retweeting out of\Nendorsement or out of ... out of ...
Dialogue: 0,0:53:57.21,0:53:59.42,Default,,0000,0000,0000,,Mic ?: Spite.\NBeyer: ... out of spite, that's right.
Dialogue: 0,0:53:59.42,0:54:03.63,Default,,0000,0000,0000,,That's the logic, why we decided to use\Nmentions and Retweets.
Dialogue: 0,0:54:03.63,0:54:05.24,Default,,0000,0000,0000,,Mic ?: Thank you.\NHerald: Another question
Dialogue: 0,0:54:05.24,0:54:09.18,Default,,0000,0000,0000,,from the internet?\NSignal: Yes. Luke23 is asking: Do you
Dialogue: 0,0:54:09.18,0:54:12.96,Default,,0000,0000,0000,,think that the window of commonly\Nacceptable ideas, the so-called Overton
Dialogue: 0,0:54:12.96,0:54:21.50,Default,,0000,0000,0000,,window, was shifted to the right by the\Nideas of the AfD echoed in the media?
Dialogue: 0,0:54:21.50,0:54:25.89,Default,,0000,0000,0000,,Beyer: That's a good question. That's a\Ngood question. Something that comes to
Dialogue: 0,0:54:25.89,0:54:30.70,Default,,0000,0000,0000,,mind here, is that media use is\Nepiphenomenal - you're sort of
Dialogue: 0,0:54:30.70,0:54:38.13,Default,,0000,0000,0000,,likely.. but the question is, like: Do you\Nthink.. does something happen in you,
Dialogue: 0,0:54:38.13,0:54:42.25,Default,,0000,0000,0000,,because you use a certain media outlet, or\Ndo you use a certain media outlet, because
Dialogue: 0,0:54:42.25,0:54:48.00,Default,,0000,0000,0000,,something happened in you ?\NFrom the sense that I got, I would say that
Dialogue: 0,0:54:48.00,0:54:53.53,Default,,0000,0000,0000,,the degree to what is.. what is acceptable,\Ndefinitely was shifted over the course of
Dialogue: 0,0:54:53.53,0:54:57.74,Default,,0000,0000,0000,,this campaign, that all of a sudden we're\Nquestioning, if remembering the Holocaust
Dialogue: 0,0:54:57.74,0:55:03.77,Default,,0000,0000,0000,,should be something that is at the heart\Nor very close to German identity.
Dialogue: 0,0:55:03.77,0:55:08.81,Default,,0000,0000,0000,,That's something that a political\Nscientist would have never expected, that
Dialogue: 0,0:55:08.81,0:55:15.89,Default,,0000,0000,0000,,this cleavage can be opened up again in a\Nway that is so potent as it did now. So it
Dialogue: 0,0:55:15.89,0:55:22.79,Default,,0000,0000,0000,,definitely did something to the overall\Ndiscourse in Germany. Whereas that is an
Dialogue: 0,0:55:22.79,0:55:30.58,Default,,0000,0000,0000,,effect of media reporting on the AFD,\Nwould require us to use something like
Dialogue: 0,0:55:30.58,0:55:35.93,Default,,0000,0000,0000,,this.. the sentiment analysis, to actually\Ndetermine how the media talked about which
Dialogue: 0,0:55:35.93,0:55:42.10,Default,,0000,0000,0000,,aspect of the AFD agendas.
Dialogue: 0,0:55:42.10,0:55:47.09,Default,,0000,0000,0000,,Herald: I can see some movement behind\Nmicrophone number 8. I'm sorry. {\i1}laughs{\i0}
Dialogue: 0,0:55:47.09,0:55:51.91,Default,,0000,0000,0000,,Mic 8: Thank you very much. Thank you for\Nyour work, I still do have a critical
Dialogue: 0,0:55:51.91,0:55:57.93,Default,,0000,0000,0000,,question. Basically, the things you showed\Nis something like we all know, yeah? We
Dialogue: 0,0:55:57.93,0:56:03.03,Default,,0000,0000,0000,,could see this happening last year, and so -\NI mean this year, in the last election. So
Dialogue: 0,0:56:03.03,0:56:08.71,Default,,0000,0000,0000,,I am wondering now, whether the method you\Nused, which was basically focusing on
Dialogue: 0,0:56:08.71,0:56:15.82,Default,,0000,0000,0000,,quantity, is in a sort of mirroring what\Nwas happening. And I'm wondering if you
Dialogue: 0,0:56:15.82,0:56:22.34,Default,,0000,0000,0000,,would work.. keep working on it. Like, you\Nused buzzwords and you used "the media"
Dialogue: 0,0:56:22.34,0:56:31.00,Default,,0000,0000,0000,,instead of, like, narrowing it down, or more..\Nusing more specific questions and I was
Dialogue: 0,0:56:31.00,0:56:37.32,Default,,0000,0000,0000,,wondering, if you have these results now and\Nyou have proof for them? What are your next
Dialogue: 0,0:56:37.32,0:56:44.27,Default,,0000,0000,0000,,questions and how can you continue to use\Nthese.. the data you have, to make it more
Dialogue: 0,0:56:44.27,0:56:50.50,Default,,0000,0000,0000,,specific, so we can really have some outcome\Nand some conclusions coming from this?
Dialogue: 0,0:56:50.50,0:56:52.87,Default,,0000,0000,0000,,Beyer: It's a absolutely wonderful\Nquestion.
Dialogue: 0,0:56:52.87,0:56:58.02,Default,,0000,0000,0000,,Of course, we thought about using\Nthis data further down the line. We.. our
Dialogue: 0,0:56:58.02,0:57:04.37,Default,,0000,0000,0000,,initial plan was, to connect this not just\Nwith salience data that we derive from
Dialogue: 0,0:57:04.37,0:57:07.71,Default,,0000,0000,0000,,Google searches.\NWe also have Facebook data that we
Dialogue: 0,0:57:07.71,0:57:11.98,Default,,0000,0000,0000,,collected, that we wanted to look into,\Nbut there.. it's a bit challenging, to
Dialogue: 0,0:57:11.98,0:57:19.78,Default,,0000,0000,0000,,actually analyze comments in depth onto\Nlanguage, because language tends to be way
Dialogue: 0,0:57:19.78,0:57:26.92,Default,,0000,0000,0000,,more fluid and you have certain problems\Nwith selection and self-selection. So you
Dialogue: 0,0:57:26.92,0:57:30.48,Default,,0000,0000,0000,,really, really have to be careful to cross-\Nconnect, which person that comments on
Dialogue: 0,0:57:30.48,0:57:38.21,Default,,0000,0000,0000,,Facebook is the same person and thus, if\Nyou only do quantitative stuff, would
Dialogue: 0,0:57:38.21,0:57:44.21,Default,,0000,0000,0000,,appear disproportionally. As I mentioned,\Nwe have also collected data from far-right
Dialogue: 0,0:57:44.21,0:57:52.49,Default,,0000,0000,0000,,blogs, from "news" blogs, that very\Nactively endorsed the AFD and their topics
Dialogue: 0,0:57:52.49,0:57:59.11,Default,,0000,0000,0000,,and so we're planning to pull this into\Nthe analysis along with data from the
Dialogue: 0,0:57:59.11,0:58:04.17,Default,,0000,0000,0000,,German Longitudinal Election Study, where\Nin this time frame, that we surveyed, in the
Dialogue: 0,0:58:04.17,0:58:10.80,Default,,0000,0000,0000,,data, each day 100 people in Germany were\Ncalled up and asked about their feelings
Dialogue: 0,0:58:10.80,0:58:16.25,Default,,0000,0000,0000,,toward specific parties and actors. So we\Nactually have day-by-day data, once it
Dialogue: 0,0:58:16.25,0:58:21.73,Default,,0000,0000,0000,,comes out, on how people.. what people\Nthought about those actors. So we're
Dialogue: 0,0:58:21.73,0:58:28.02,Default,,0000,0000,0000,,planning to pull that in, as a more\Nreliable measure for salience.
Dialogue: 0,0:58:28.02,0:58:30.79,Default,,0000,0000,0000,,Herald: Thank you very much. I'm very\Nsorry, but time's up, so there will be no
Dialogue: 0,0:58:30.79,0:58:35.03,Default,,0000,0000,0000,,more questions right now in front of the\Naudience. Alexander Beyer, thank you very
Dialogue: 0,0:58:35.03,0:58:38.03,Default,,0000,0000,0000,,much. A warm applause, please.\N{\i1}applause{\i0}
Dialogue: 0,0:58:38.03,0:58:40.86,Default,,0000,0000,0000,,Beyer: Thank you.\N{\i1}applause continues{\i0}
Dialogue: 0,0:58:40.86,0:58:45.95,Default,,0000,0000,0000,,{\i1}postroll music{\i0}
Dialogue: 0,0:58:45.95,0:59:03.00,Default,,0000,0000,0000,,subtitles created by c3subtitles.de\Nin the year 2019. Join, and help us!